--> Abstract: A Neural Network Application in Biostratigraphy, by J. Yang-Logan and J. M. Hornell; #90987 (1993).

Datapages, Inc.Print this page

YANG-LOGAN, JUDY, and JAMES M. HORNELL, Phillips Petroleum Company, Bartlesville, OK

ABSTRACT: A Neural Network Application in Biostratigraphy

We have developed an algorithm to train a neural network to make age and paleoenvironmental determinations from biostratigraphic data. Neural network computing is a new, adaptive computing method. The neural network is trained from examples for which age and paleoenvironment have already been determined. After training the neural network makes age and paleoenvironmental determinations for new well data.

The examples used for training include areas from California, the North Sea, North Africa, and the Gulf of Mexico. The age of the examples ranges from Mesozoic to Tertiary.

The neural network makes more specific determinations than was thought to be possible with the available data. Less than a minute is required for the neural network to make age and paleoenvironment determinations for an entire well. The neural network was trained with species, and it was also trained with only the genera. The neural network was trained with counts, and it was trained with only presence and absence. Preliminary results indicate that paleobathymetry decisions for the North Sea areas are similar in these cases, species, genera only, counts, and presence and absence only. Neural network age decisions are more complicated than paleoenvironmental decisions.

This new approach will shorten the time required for biostratigraphic analysis. It will provide better correlations by increasing accuracy and consistency.

AAPG Search and Discovery Article #90987©1993 AAPG Annual Convention, New Orleans, Louisiana, April 25-28, 1993.